V2 Walk-Through a Stitched Image

Abstract

In imaging devices, there is a trade-off between the images (snapshots) and video because of the limitation in storage capacity. Video clips need more storage space compared to images. This motivated to generate the video from a single stitched 2D image rather than storing video clips. Humans analyze varieties of cues from a single image and act accordingly, unlike robots. Creating a vertical virtual walk-through a Single stitched large sized image is proposed in two folds: In the first phase, the input images are stitched to form a single large sized image of a scene and in the second phase the virtual walk through the stitched image is rendered. The algorithm is tested on a set of nearly 400 images obtained from both indoor and outdoor scenes. The presented approach is computationally efficient and has been verified extensively by experiments. Since the proposed algorithm is completely automated it is amenable for implementation on robotic platforms.